The existing phase calculation must obtain three or more images of different phases, for example, 0, Π/2 and Π, and then phase data is calculated according to the three images. Such detection is relatively easy to implement for stationary objects, while for moving objects, as it is necessary to set at least three channels of optical elements for image detection, the structure is complicated and the cost is also very high.